Implications of physical symmetries in adaptive image classifiers

Symmetries like rotational invariance and reflection in image classifiers reduce parameters. This improves neural network training efficiency and generalization without hindering object distinction.